The Noise Controller

Noise in biological and chemical systems is traditionally treated as a problem — random fluctuations that blur signals, degrade performance, and must be filtered out or averaged over. The machinery of life, in this view, works despite its noise.

De Giuli (arXiv:2503.15670) inverts the picture: noise equals control. In the weak-noise limit, the most probable trajectory of a stochastic system between two states is mathematically identical to the optimal control strategy that minimizes a specific action functional. The noise is not a perturbation on the dynamics — it is the control variable.

For Langevin dynamics, this is a formal equivalence: the stochastic force term plays exactly the role of the control input in the optimal control formulation. For general Markov jump processes — chemical reaction networks, electronic circuits, gene regulatory networks — the Doi-Peliti response field serves as the control variable. In every case, the “noise” is what steers the system optimally between states.

This framework explains several counterintuitive phenomena. Stochastic resonance — where adding noise to a signal improves its detection — is not a paradox but a natural consequence: the noise is doing the optimal thing, which happens to be amplifying the signal. Brownian ratchets extract directed motion from thermal fluctuations not by exploiting a loophole in thermodynamics but by implementing an optimal control strategy using noise as the control input.

The structural insight: the distinction between noise and control is not physical but perspectival. The same stochastic force that looks like random perturbation from outside looks like optimal steering from inside. Biological systems don’t work despite their noise. They work with it — and in the mathematical sense, they work through it.

De Giuli, “Noise equals control,” arXiv:2503.15670 (2026).